Memory

Persistent agent memory that works across sessions

The --memory flag and memory command enable persistent memory for agents, allowing them to remember context across sessions.

Install requirement: most praisonai memory subcommands (show, list, add, search, clear) require pip install praisonai (the full wrapper). On a standalone pip install praisonai-code install they exit 1 with memory requires the full wrapper. Install the full wrapper: pip install praisonai — a single-line hint, no Rich traceback (PR #2854). The native memory learn subcommand stays standalone-safe (see the Tip below).

memory learn (and its status, show, add, search, clear sub-subcommands) is a native subcommand — it runs in-process via praisonaiagents.memory.LearnManager and works on a standalone pip install praisonai-code install without the wrapper.

Quick Start

Show Memories

praisonai memory show

Use Memory Flag

# Enable memory for agent
praisonai "count to 5" --memory

Usage

Enable Memory

# Enable memory for agent (persists across sessions)
praisonai "My name is John" --memory

# Memory with user isolation
praisonai "Remember my preferences" --memory --user-id user123

Expected Output:

🧠 Memory enabled

╭─ Agent Info ─────────────────────────────────────────────────────────────────╮
│  👤 Agent: DirectAgent                                                       │
│  Role: Assistant                                                             │
│  Memory: Enabled (user: user123)                                            │
╰──────────────────────────────────────────────────────────────────────────────╯

╭────────────────────────────────── Response ──────────────────────────────────╮
│ Nice to meet you, John! I'll remember your name for future conversations.   │
╰──────────────────────────────────────────────────────────────────────────────╯

💾 Memory saved: name="John"

Memory Commands

Show Memory

# Show 10 most recent memories (default)
praisonai memory show

# Change how many are shown
praisonai memory show --limit 25
praisonai memory show -l 25

# Isolate by user
praisonai memory show --user-id user123

Output:

╭─ Memory Statistics ──────────────────────────────────────────────────────────╮
│  Short-term: 5 items                                                        │
│  Long-term: 12 items                                                        │
│  Entities: 3 items                                                          │
│  Episodic: 8 items                                                          │
╰──────────────────────────────────────────────────────────────────────────────╯

List Memories

memory list is an alias for memory show — same output, added for parity with tools list, session list, and config list.

# List memories (default: 10 most recent)
praisonai memory list

# Show a specific number
praisonai memory list --limit 25

# Isolate by user
praisonai memory list --user-id user123 --limit 5

Use -l as a short form: praisonai memory list -l 25.

memory show and memory list accept the same flags:

Flag	Type	Default	Description
--user-id	str	None	User ID for memory isolation
--limit, -l	int	10	Number of memories to show

Add Memory

praisonai memory add "User prefers Python"

Search Memory

praisonai memory search "Python"

Clear Memory

# Clear short-term memory (the default target)
praisonai memory clear

# Clear everything
praisonai memory clear all

Behaviour change (PraisonAI PR #4946). praisonai memory clear now prompts before deleting. Automation must pass --force (or -f). Previously the command advertised --force but ignored it, and clear all failed with Got unexpected extra argument.

memory clear and memory clear all prompt for confirmation, matching the memory learn clear sibling command’s flow:

$ praisonai memory clear all </dev/null
Clear ALL memories? This cannot be undone. [y/N]: Cancelled.   # EXIT=1, memory intact

$ praisonai memory clear all --force
✅ All memory cleared                                          # EXIT=0

Target	Clears	Notes
short	Short-term memory only	Default when no target is given
all	Everything	Now a valid target argument and works end-to-end

Flag	Short	Description
--force	-f	Skip the confirmation prompt (required for scripts/CI)
--user-id		Scope the clear to a single user’s memories

An unknown target (anything other than short/all) exits 2; aborting the prompt exits 1 with nothing cleared.

Session Management

# Save session
praisonai memory save my_session

# Resume session
praisonai memory resume my_session

# List saved sessions
praisonai memory sessions

Checkpoints

# Create checkpoint
praisonai memory checkpoint

# Restore checkpoint
praisonai memory restore <checkpoint_id>

# List checkpoints
praisonai memory checkpoints

Help

praisonai memory help

Memory Types

Type	Description	Persistence
Short-term	Rolling buffer of recent context	Auto-expires
Long-term	Important facts sorted by importance	Persistent
Entity	People, places, organizations	Persistent
Episodic	Date-based interaction history	Persistent

How It Works

1. Capture: Agent extracts important information from conversations
2. Categorize: Information is sorted into memory types
3. Store: Memories are persisted to storage
4. Inject: Relevant memories are injected into future conversations

Storage Options

Option	Dependencies	Description
memory=True	None	File-based JSON storage (default)
memory="file"	None	Explicit file-based storage
memory="sqlite"	Built-in	SQLite with indexing
memory="chromadb"	chromadb	Vector/semantic search

Programmatic Usage

from praisonaiagents import Agent
from praisonaiagents import FileMemory

# Enable memory with a single parameter
agent = Agent(
    name="Personal Assistant",
    instructions="You are a helpful assistant that remembers user preferences.",
    memory={"user_id": "user123"}  # Enables memory with user isolation
)

# Memory is automatically injected into conversations
result = agent.start("My name is John and I prefer Python")
# Agent will remember this for future conversations

Advanced Features

from praisonaiagents import FileMemory

memory = FileMemory(memory={"user_id": "user123"})

# Session Save/Resume
memory.save_session("project_session", conversation_history=[...])
memory.resume_session("project_session")

# Context Compression
memory.compress(llm_func=lambda p: agent.chat(p), max_items=10)

# Checkpointing
memory.create_checkpoint("before_refactor", include_files=["main.py"])
memory.restore_checkpoint("before_refactor", restore_files=True)

# Slash Commands
memory.handle_command("/memory show")
memory.handle_command("/memory save my_session")

Best Practices

Use --user-id to isolate memory per user in multi-user applications.

Memory storage grows over time. Use memory clear periodically to manage storage.

Do	Don’t
Use user isolation for multi-user apps	Share memory across users
Clear short-term memory regularly	Let memory grow unbounded
Use checkpoints before major changes	Skip backups
Search before adding duplicates	Add redundant memories

Related

• Memory Concept
• Auto Memory CLI
• Session CLI